Claims routing sounds simple: send straightforward files to generalists, send complex files to specialists. In practice, the routing decision is made under time pressure with incomplete information, and the consequences of a wrong assignment compound across the life of the file. A complex commercial casualty claim routed to a standard auto handler will be mishandled for days before the error surfaces -- and by then, coverage positions may have been taken, first contact may have been made, and the file has a history that is hard to unwind.
What makes a file "complex"
Complexity in claims is not simply a function of dollar amount, although large-dollar files are statistically more likely to be complex. The defining characteristics of a complex file are: disputed liability or coverage, multiple parties or claimants, bodily injury or fatality, commercial policyholder with non-standard endorsements, potential for subrogation recovery, litigation indicators in the FNOL narrative, or jurisdictional complications.
A $50,000 straightforward property replacement claim for a homeowner with a clean policy history is not complex. A $9,000 auto claim involving a commercial vehicle, a third-party claimant alleging back injury, and an employer that may bear some liability is complex regardless of the reported amount. The amount-based routing rules that many carriers use by default frequently send this second type of claim to the wrong queue.
The cost of misrouting
When a complex file lands in a standard adjuster queue, two things typically happen. Either the handler realizes the complexity quickly and escalates -- creating a rework cycle that adds 2 to 5 days to file handling and forces the handler to document their initial work before transferring -- or the handler does not realize the complexity and proceeds, making decisions that are correct for a simpler version of the claim but wrong for the actual file in front of them.
The second scenario is worse and is more common than claims managers acknowledge. Handlers under volume pressure tend to process files as they find them. Escalation requires time and documentation. When the queue is long, the path of least resistance is to keep moving. This is not a failure of individual handlers -- it is a predictable consequence of routing systems that put complex files in front of people who are not equipped to handle them.
Quantifying the cost of misrouting is difficult because it is embedded in average cycle times, reserve development, and litigation rates rather than a direct line item. A carrier that tracks escalation rates by initial queue assignment will find the signal. Files that escalate after initial handler contact close 40 to 80% slower than files routed correctly from the start, in typical carrier data.
What routing logic actually requires
Effective routing logic at FNOL requires the following inputs: loss type classification (with subclass), coverage line and endorsements, third-party involvement flag, bodily injury indicator, commercial vs. personal policy flag, reported amount, and any litigation or represented-claimant indicators from the narrative. These inputs drive a complexity score that maps to a routing tier.
Rule-based routing systems can handle the structured inputs adequately -- coverage line and commercial/personal flag are typically available in policy data. What they cannot handle is the narrative inputs: the sentence in the loss description that says "claimant has retained an attorney," the mention of a delivery vehicle that implies commercial use, or the description of symptoms that suggests soft-tissue injury even when the FNOL does not explicitly categorize it as bodily injury.
Language extraction from the FNOL narrative is where automated routing adds the most value over rule-based systems. The difference between routing 70% of files correctly and routing 92% of files correctly is almost entirely in the handling of narrative signals that structured rules cannot see.
Complexity scoring in practice
The Adjustsage complexity score runs from 0 to 100 and is computed from a weighted combination of structured and extracted features. A simple homeowner property claim with no third-party involvement, no bodily injury indicators, and a reported amount under $8,000 scores below 25 and proceeds to the straight-through processing queue. A personal auto claim with bodily injury indicators from the narrative, a third-party claimant mentioned by name, and a commercial endorsement on the policy scores above 75 and routes to a senior adjuster queue with the complexity flags surfaced in the assignment notification.
Carriers configure their own routing thresholds during onboarding. A carrier with a specialist litigation unit may set the threshold for that queue at 65; a carrier without one may set it at 85 and route high-complexity files to senior generalists. The scoring logic is consistent; the threshold configuration is carrier-specific.
Human override is not a failure mode
No automated routing system is right 100% of the time. The override workflow matters as much as the accuracy of the initial routing. Handlers should be able to escalate or de-escalate files with minimal friction, and every override should be logged with the handler's stated reason. This log is valuable for two purposes: it provides training data for improving the model, and it identifies systematic errors in routing logic.
Carriers that treat override rates as a failure metric miss the point. A system that routes 90% of files correctly and makes the 10% that need human review easy to catch and correct is better than a system that routes 95% correctly but makes overrides painful -- because the painful override system accumulates errors in the 5% it misses rather than surfacing them for correction. The goal is not to minimize overrides; it is to minimize mishandled claims from first contact to close.
Integration with adjuster queue management
Routing decisions are only useful if they land in an adjuster's actual workflow. The Adjustsage routing output is designed to push directly into existing claims management system queues via the Production tier webhook integration. When a file routes to a senior adjuster queue, the CMS receives a structured payload with the complexity score, the flags that drove it, and the extracted FNOL fields -- so the adjuster opening the file sees a pre-populated intake record with the relevant signals already highlighted.
This integration eliminates the copy-paste step that handlers currently perform between intake tools and CMS. It also means the routing rationale is captured in the CMS record from day one, which simplifies quality review and supports any later disputes about how the file was initially assessed.